Announcement

Collapse
No announcement yet.
X
  • Filter
  • Time
  • Show
Clear All
new posts

  • Panel data: Heteroscedasticity and Autocorrelation => Newey West HAC Standard errors?

    Hello,

    i have the following problem. My dataset is panel data like the extract below shows.
    Obs_ year id conm age patents cites
    1 1976 32 A & E PLASTIK PAK INC 1 2 21
    2 1977 32 A & E PLASTIK PAK INC 2 3 31
    3 1976 370 A.B.A. INDUSTRIES INC 1 0 0
    4 1977 370 A.B.A. INDUSTRIES INC 2 0 0
    5 1978 370 A.B.A. INDUSTRIES INC 3 0 0
    6 1979 370 A.B.A. INDUSTRIES INC 4 0 0
    7 1980 370 A.B.A. INDUSTRIES INC 5 0 0
    8 1981 370 A.B.A. INDUSTRIES INC 6 0 0
    9 1979 774 ABKCO INDUSTRIES INC 1 0 0
    10 1980 774 ABKCO INDUSTRIES INC 2 0 0
    11 1981 774 ABKCO INDUSTRIES INC 3 0 0
    12 1982 774 ABKCO INDUSTRIES INC 4 0 0
    13 1983 774 ABKCO INDUSTRIES INC 5 0 0
    Concerning the model structure I ran a Hausman test with the result that a fixed effects model should be applicable.
    After that I ran several other Regression diagnostics : Homoscedasticity, autocorrelation and multi-collinearity.
    The tests revealed that both heteroscedasticity and autocorrelation are present in my model.
    Reading some other papers left me with the following solution: Newey West Standard Errors.

    My regression equation now should look like this:

    newey dep indep1 indep2 indep3 ... i.year i.industry, lag(X) force

    The force command is needed because of the panel data form. My question is now how to compute the lag?
    Should i use X=1 or is there a standard variable?

    Best regards
    Nic
Working...
X